Six things, all found by using the deployed page.
**The icons were never images.** `everness.info/...` answers 200 with
`content-type: image/webp` for paths that do not exist, returning the SPA's
HTML shell, so all 72 downloaded "icons" were HTML documents named .webp and
every character silently rendered as initials. The assets live on
`api.everness.info/data/assets` instead, and the fetcher now verifies RIFF/WEBP
magic bytes and fails the build rather than warning. 95 files: 23 avatars at
256px, 23 portraits, 49 Arc icons.
**Stats read as the game writes them.** `prune-gamedata` harvests every
`{ id_stats, name, bShowPercent }` pair out of the mirror, so "CRIT Rate" is
the game's own wording rather than someone's typing, and values render as 12.5%
instead of 0.125. Ids the mirror never names are filled in by
`domain/statlabels.ts`, each with its reason and marked derived in the UI.
Two of everness's own percent flags are overridden: it marks `HealUp` and
`DamageUpNatureBase` flat while shipping them as fractions in exactly the range
of the bonuses it does mark as percentages, which printed "Healing Bonus 0.1".
**The build is legible without hovering.** Each piece carries its legend number
and colour, and the list beside the board states shape, type, level and every
stat outright; the cartridge sits in its own box below, since it occupies no
cells. Hover-only tooltips meant the information existed but could not be
scanned or compared. Shapes are drawn from the same cell lists the solver packs
with, so an icon cannot drift from its geometry.
**Owner groups have numbers, not hashes.** "group 0000000088" told the player
nothing; groups are now numbered stably and an unnamed one links to where it
gets named.
Also: import/export moved to its own Data tab with a library summary and the
to-do list, and the layout is denser throughout.
152 tests, tsc clean, both build targets clean.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The homelab has no rsync, so the tree goes over as a tar stream, and the
destination is wiped first - a stale content-hashed chunk left behind would be
served forever, since /assets/* is cached as immutable.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Found by running the app in a browser for the first time.
- Undo was only reachable from the team tab, and only after a team solve had
finished, so an equip made from the character view had no way back. Both
views now show it whenever there is something to undo.
- The owner-group naming dropdown listed all 23 espers from the game database
rather than the characters the capture says you own. Two espers are both
published as "Zero", so the list held two identical entries and picking the
wrong one misattributed a whole loadout; conversely a character the game
database has never heard of had no entry at all and its group could not be
named. Options now come from the captured roster, labelled with the esper
name and disambiguated by codename when two owned characters share one.
- An inline SVG favicon, so the page stops requesting one that does not exist.
Inline rather than a file because the single-file build must make no
requests at all.
Verified headless against the real 819-item export: import writes 819 items,
20 characters and 99 equipment rows to IndexedDB; a solve runs in a real
Worker and scores 0.6690 for Zankou, matching the Node bench; equip writes 8
app rows and displaces 7; undo restores all 99 imported rows; state survives
a reload. The file:// build renders, imports, solves inline and issues no
network request.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Client-side optimizer for Neverness to Everness console builds. Imports a
gear export from nte-history-exporter, keeps it in IndexedDB, and solves for
the best module placement per character or across a team.
- domain/: pure ports of the Python research scripts (shapes, cartridges,
board tiling, stats, scoring, set bonuses, Arc effects, sheet prediction).
No DOM, no storage imports, so they stay worker-safe and Node-testable.
- solver/: single-character solve with branch and bound, plus a three-phase
team solve (portfolio, leximin, column generation) over disjoint item sets.
Runs in a Worker when one can be constructed, inline otherwise.
- db/: three storage tiers selected by use rather than feature detection,
fail-closed import validation, and undo that restores both sides of a steal.
- ui/: React 19 views for items, characters and teams; the board is drawn as
the deliverable rather than a score.
- Two build targets: a hosted bundle and a single self-contained nte.html
with no external requests.
Values the model cannot measure are reported as unavailable rather than
guessed: cartridge set bonuses, the level 80 base-stat multiplier, and the
distinction between "optimal for this packing" and globally optimal.
145 tests, tsc clean, both build targets clean.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
